无网络环境下如何从Anaconda base环境复制指定包到新环境？

Hey there! This is such a common pain point when working offline with Conda—no need to waste time copying every single package from base. Let’s break down the best methods to transfer exactly the 10-20 packages you need, no internet required.

First, Understand Conda’s Local Cache #

Conda automatically stores every package you’ve ever installed (including those in base) in a local cache directory. On most systems:

• Linux/macOS: ~/anaconda3/pkgs/
• Windows: C:\Users\YourUsername\anaconda3\pkgs

All the .conda or .tar.bz2 files here are the installers for your base packages—we’ll leverage this instead of downloading anything new.

Method 1: Use Conda’s --offline Flag (Recommended for Conda-Installed Packages) #

This is the simplest way if your target packages were installed via Conda in the base environment.

1. List base environment packages to get versions:
   Run this to see all packages in base, note the exact version numbers for the ones you want:

   conda list -n base
   

   For example, you might see numpy 1.26.0 py311h59b608b_0—you’ll need numpy=1.26.0 for the install command.

2. Activate your new environment:

   conda activate your-new-env-name
   

3. Install specific packages offline:
   Use the --offline flag to force Conda to only use its local cache. List your target packages with their versions:

   conda install --offline numpy=1.26.0 pandas=2.1.3 matplotlib=3.8.0
   

   Conda will pull all these packages directly from the pkgs cache directory—no internet needed.

Method 2: Use a Package List File #

If you have a long list of packages, it’s easier to create a text file with the package names and versions, then install from that.

1. Create a required-packages.txt file:
   Add each package and its version on a new line, like this:

   numpy=1.26.0
   pandas=2.1.3
   scipy=1.11.3
   seaborn=0.13.0
   

2. Install from the file offline:
   Activate your new environment, then run:

   conda install --offline --file required-packages.txt
   

Method 3: For Pip-Installed Packages in Base #

If some of your target packages were installed via pip in base (not Conda), you can use Pip’s local cache to install them offline.

1. Locate Pip’s cache directory:

   • Linux/macOS: ~/.cache/pip/wheels/
   • Windows: C:\Users\YourUsername\AppData\Local\pip\Cache\wheels

2. Install from local wheels:
   Activate your new environment, then use --no-index (skip PyPI) and --find-links (point to local cache directories):

   pip install --no-index --find-links=~/anaconda3/pkgs --find-links=~/.cache/pip/wheels numpy pandas
   

   This tells Pip to look for package wheels in both Conda’s cache and its own local cache.

Quick Note: There’s No --use-locals Flag, But... #

You’re right that there’s no exact --use-locals parameter for Conda, but the --offline flag achieves exactly what you want—it restricts Conda to only use packages already present on your machine.

Troubleshooting Tip #

If a package isn’t found in the cache, it’s likely that the base environment’s version was compiled from source (not downloaded as a pre-built package). In that case, you’d need to manually copy the package’s files from base/lib/pythonX.X/site-packages/ to your new environment’s site-packages/ directory, but this is rare for most common packages.

内容的提问来源于stack exchange，提问作者Mikhail_Sam